Queuing Model Based Edge Placement for Work Offloading in Mobile Cloud Networks

Tai-Lin Chin

doi:10.1109/ACCESS.2020.2979479

2020-01-01

Tai-Lin Chin

Yeong-Sheng Chen *

Kun-Yu Lyu

DOI：10.1109/ACCESS.2020.2979479

原文链接

查看原文

摘要

En 中文

Recently, many useful services have been provided through wireless mobile devices. One of the critical issues is that mobile devices are generally built with limited computing capability, and thus may not be able to handle the complexity of many advanced services. This problem can be mitigated by offloading certain jobs from the mobile devices to the edge or cloud servers. However, the locations of the edge servers and the allocation of the offloading jobs to the edge servers can significantly affect the traffic generated in the network. Accordingly, this study conducts an investigation into the edge server placement and work allocation strategy with a focus on minimizing the total traffic load so as to achieve green communications in the mobile cloud networks. Mobile devices are assumed to be attached to the nearby node and they may send work offloading requests to the attached nodes. Meanwhile, edge servers are deployed at carefully selected nodes. The offloading jobs received by a node are assigned to a selected edge server. If the edge server is busy, the work will be forwarded to the central cloud server. Otherwise, it is processed locally in the edge server. A queuing model is adopted to evaluate the job forwarding probability at the edge servers. The problem is formulated as an integer programming problem. Two novel heuristic schemes are proposed, namely the Set-by-Set algorithm and the -clustering algorithm. The simulation results showed that the -clustering algorithm consistently outperforms both the Set-by-Set algorithm and the Density-Based-Clustering (DBC) algorithm presented in the literature in terms of a lower total traffic load within the network.

Keyword:

Servers

Cloud computing

Mobile handsets

Resource management

Linear programming

Heuristic algorithms

Clustering algorithms

Edge computing

cloud networks

work offloading

green communications